CN-114987511-B - Method for simulating human driving behavior to train motion controller based on neural network

CN114987511BCN 114987511 BCN114987511 BCN 114987511BCN-114987511-B

Abstract

A number of variations may include a method of training a neural network vehicle motion controller that more closely replicates how a human would drive a vehicle using intuitive vehicle dynamics variables and predicts parameters to determine how the motion controller should communicate steering angle, throttle, and interrupt inputs to the vehicle to navigate the vehicle.

Inventors

O. Kraft

Assignees

大陆汽车系统公司
操纵技术IP控股公司

Dates

Publication Date: 20260505
Application Date: 20210506
Priority Date: 20210301

Claims (3)

1. A method of training a neural network, comprising causing a human driver to drive a vehicle on a test runway at a first rate for a first driving characteristic and using a plurality of sensors and one or more modules or computing devices to determine a current state of the vehicle at a plurality of points in time using at least one of yaw, speed, lateral acceleration, longitudinal acceleration, yaw rate, speed, steering wheel angle, and steering angle targets, wherein for the first driving characteristic the human driver turns in a rapid or aggressive manner and accelerates or decelerates in a aggressive or rapid manner; and determining predictions made by a driver, namely at least one of path geometry in front of the vehicle including an X-direction and a Y-direction, lateral deviation of the vehicle from an intended path, heading deviation of a current heading of the vehicle from the intended path, curvature of a future trajectory, and target speed, and generating input data based on the determinations, and communicating the input data to a neural network to simulate human driving behavior and generate output data from the neural network, and communicating the output data to an autonomous driving vehicle module constructed and arranged to drive the vehicle without human input for at least a period of time; Wherein the method further comprises causing a human driver to drive the vehicle on the test runway at a second rate for a second driving characteristic and using a plurality of sensors and one or more modules or computing devices to determine a current state of the vehicle at a plurality of points in time using at least one of yaw, speed, lateral acceleration, longitudinal acceleration, yaw rate, velocity, steering wheel angle and steering angle targets, wherein for the second driving characteristic the human driver turns in a less aggressive manner and accelerates or decelerates in a slower manner than the first driving characteristic, and determining predictions made by the driver, i.e., at least one of path geometry in front of the vehicle including X-direction and Y-direction, lateral deviation of the vehicle from an intended path, heading deviation of the vehicle current from an intended path, curvature of the future path and target speed, and generating input data in accordance with the determination, and communicating the input data to a neural network to simulate human behavior and generate output data from the neural network, and arranging the output data to the vehicle in an autonomous manner and communicating the output data to at least one of the driver module, wherein the vehicle is configured to drive at least one of the second rate and the human driver is small for a period of time and the second rate is small, wherein the prediction made by the driver is determined by the driver is at least one of the speed of the vehicle's current and the vehicle's own driving rate Wherein the method further comprises causing a human driver to drive the vehicle on the test runway at a third rate for a third driving characteristic and accelerating or decelerating in a slower manner than the second driving characteristic and using a plurality of sensors and one or more modules or computing devices to determine a current state of the vehicle at a plurality of points in time using at least one of yaw, speed, lateral acceleration, longitudinal acceleration, yaw rate, velocity, steering wheel angle and steering angle targets, wherein for the third driving characteristic the human driver turns slower and less severely and accelerates or decelerates in a slower manner than the second driving characteristic, and determining predictions made by the driver, i.e. at least one of path geometry in front of the vehicle including X-direction and Y-direction, lateral deviation of the vehicle from an intended path, heading deviation of the vehicle current from an intended path, curvature of the future path and target speed, and communicating the input data to a neural network to simulate human driving behavior and generate output data from the neural network, and communicating the output data to the vehicle at least within an autonomous driving module, wherein the vehicle is not required to be configured for a small period of time and wherein the vehicle is driven at least, Wherein the trained neural network is constrained downstream to remain within safe operating limits.
2. A system of trained neural networks constructed and arranged to produce output data, the neural networks having been trained by the method of claim 1.
3. A method comprising training the system of claim 2 with a predetermined neural network model architecture, the method comprising determining an inherent uncertainty of the set of training data and an uncertainty within the predetermined neural network model architecture prior to feeding the set of training data to the neural network, thereby causing data preprocessing to determine homodyne and homodyne uncertainties, and using them as inputs to allow the neural network to understand and learn how the inputs are spread in driving space, and learning/adjusting the mean and standard deviation associated with each network neuron of neural network weights and deviations.

Description

Method for simulating human driving behavior to train motion controller based on neural network Technical Field The field to which the disclosure generally relates includes vehicle motion controllers, and methods of making and using the same (including methods of simulating human driving behavior to train neural network-based vehicle motion controllers). Background Autonomous and semi-autonomous vehicles may use motion controllers to control longitudinal and lateral motion of the vehicle. Disclosure of Invention Many variations may include vehicle motion controllers, and methods of making and using the same (including methods of simulating human driving behavior to train neural network-based vehicle motion controllers). A number of variations may include a method of training a neural network vehicle motion controller that more closely replicates how a human would drive a vehicle using intuitive vehicle dynamics variables and predicts parameters to determine how the motion controller should communicate steering angle, throttle, and interrupt inputs to the vehicle to navigate the vehicle. Other illustrative variations within the scope of the invention will become apparent from the detailed description provided hereinafter. It should be understood that the detailed description and specific examples, while disclosing variations within the scope of the invention, are intended for purposes of illustration only and are not intended to limit the scope of the invention. Drawings Selected examples of variations within the scope of the invention will be more fully understood from the detailed description and the accompanying drawings, in which: Fig. 1 illustrates a method of training a neural network to simulate human driving behavior, which may include characterizing the current state of a vehicle, what the driver sees in terms of path geometry, and perceived errors corrected by the driver by applying steering and throttle/brake inputs. Fig. 2 is a block diagram of a trained neural network including trained parameters based on neural network architecture, where X1 is a vector of training inputs shown in fig. 1, and Y1 is a vector of control parameters that are sent to actuators to control lateral and longitudinal motion of the vehicle. Fig. 3 is a block diagram illustrating a method of training a neural network. Detailed Description The following description of the variations is merely illustrative in nature and is in no way intended to limit the scope, application, or uses of the invention. Many variations may include vehicle motion controllers, and methods of making and using the same (including methods of simulating human driving behavior to train neural network-based vehicle motion controllers). Various variations may include a method of training a neural network vehicle motion controller that more closely replicates how a human will drive a vehicle using an "intuitive" feel characterized by vehicle dynamics variables, and predicting parameters to determine how the motion controller should transmit steering angle, throttle, and interrupt inputs to the vehicle to navigate the vehicle. Heretofore, lateral and longitudinal vehicle motion controllers have been separate and only when control inputs were provided to the vehicle actuators inferred their effects on each other's vehicle dynamics. These types of motion control methods lend themselves to very robotic or unnatural vehicle behavior, which makes human vehicle drivers and/or passengers significantly perceived as lively and uncomfortable. In various variations, the prediction data may be used as or parameterized into a system of equations represented by a multi-order differential equation. Later, this data can be fed to the neural network in an input-output form (the input-output form is prepared in advance so that the network obtains weights and deviations that will fit the input data set as closely as possible). These weights and deviations may then be deployed as a "uniform motion controller (homogeneous motion controller)" to enable lateral and longitudinal vehicle motion control in an autonomous or semi-autonomous mode. For braking, this is also possible. The weights and deviations may be deployed as a "uniform motion controller" to implement vehicle deceleration motion control in an autonomous or semi-autonomous mode. The inputs to such a vehicle motion controller will be exactly the same as the inputs in terms of variables used in the training. But due to the nature of the general nature of the neural network, the neural network will be robotic-like for variation when compared to training data and will be able to drive forward on the desired road at the desired rate required by the path planner. Because the neural network has been trained on the same input vector based on learned behavior modeled in weights, biases, and related process uncertainties, the output of the controller will very closely match what a human would do (if the same set of inputs were